JPH1139019A - Off-line teaching method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the number of work processes using a real robot by ena bling teaching data, that can estimate the amount of correction with a high precision, to be steadily obtained. SOLUTION: In this method, an off-line teaching for correcting a robot model in an off-line teaching system is performed. In this case, a multi-point multi- attitude teaching, which can obtain the best correction calculation result from a logical robot model of a correction object, is performed in an off-line teaching device (steps S1 to S5), and teaching data obtained from these steps S1 to S5 are downloaded in a real robot of the correction object from the off-line teaching device (step S6). Next, the teaching data are corrected on the basis of a difference between an action point of the real robot, which is based on the teaching data, and a real target point (step S7). Then, teaching data after correction are uploaded to the off-line teaching device (step S8) and an estimation calculation of the amount of correction is executed by the uploaded data (step S9).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an offline teaching method capable of accurately estimating a correction amount of teaching data previously obtained by teaching a logical robot model with an offline teaching device.

[0002]

2. Description of the Related Art In recent years, in order to apply a robot to various tasks, it has become common to attach various tools to a face plate of a robot arm to cause the robot to perform a task.

[0003] Conventionally, many teaching techniques for a robot and various techniques for making various corrections to the teaching data have been proposed. For example, regarding the teaching technique for a robot, when inputting information for teaching, the information necessary for the robot operation to be used should be as accurate as required by the user, and the burden on the user should be reduced as much as possible. (Japanese Unexamined Patent Publication No. 5-2782)
No. 8).

[0004] Regarding the technique of correcting teaching data,
A method in which the teaching point correction and the teaching trajectory correction can be easily performed under the same operating conditions as at the time of playback even in a place away from the work site, thereby reducing the burden on the operator for the teaching correction work (particularly, (See Japanese Unexamined Patent Publication No. Hei 8-286726) and a method for automatically correcting the positional deviation with high accuracy for all the hitting positions (Japanese Patent Laid-Open No. 7-3256).
No. 11) or a method of correcting the position of a robot using a neural network (Japanese Patent Laid-Open No. 6-1146).
No. 9), and a method in which the operator can perform intuitively easy-to-understand correction work when correcting the conversion data of the robot operation when the corrected teaching data has an operation range abnormality. JP-A-5-28973
No. 0) Or, to eliminate the measurement on the orthogonal coordinates and to secure the absolute position accuracy of the arm type articulated robot only by the rotation angle of the arm axis, the parameters of the unknown variable array and the constant array are modified. A method in which logical operation is applied to further improve accuracy and increase operation speed
Japanese Unexamined Patent Publication No. 2-274213), a method of automatically correcting a setting error of a constant and a setting error of a tool offset which are setting data of a robot having a tool attached to a wrist of an articulated robot (see Japanese Patent No. 2520324), and the like. Has been proposed.

[0005] Regarding the setting of the tool tip point, even if the design data is not available, the tool tip point can be set in a desired posture by a simple procedure using a simple setting jig. Method (Japanese Unexamined Patent Publication No.
No. 191738) is proposed, and the use of CAD data eliminates the need for the operator to input initial setting data, work route data, and work operation data one by one, so that the amount of input by the operator can be greatly reduced. (Refer to Japanese Patent Application Laid-Open No. 8-286722), which relates to the trajectory display, when the work is retracted from the work position, the relative position between the actual tool position during operation and the work. A method for easily and accurately recognizing a relationship (Japanese Patent Laid-Open No. 8-174454)
Reference) has been proposed.

[0006]

However, in the above-mentioned conventional technique relating to off-line teaching, the work point of the robot is set as different as possible with respect to a fixed point in the space at the site where the real robot is installed. Is positioned with high accuracy, and the attitude is left to the discretion of the operator.

As described above, conventionally, since it is left to the individual judgment of the operator, there is a disadvantage that teaching data suitable for calculating the correction amount cannot always be obtained. There was also the problem of this.

The present invention has been made in consideration of such problems, and can stably acquire teaching data from which a correction amount can be accurately estimated, and reduce the number of work steps using a real robot. It is an object of the present invention to provide an offline teaching method that can be used.

[0009]

An off-line teaching method according to the present invention provides an off-line teaching method for correcting a robot model in an off-line teaching system, whereby the best correction calculation result is obtained with a logical robot model to be corrected. A first step of performing multi-point multi-position teaching using an offline teaching device, a second step of downloading the teaching data obtained in the first step from the offline teaching device to a real robot to be corrected, and the teaching A third step of correcting the teaching data based on a difference between an operating point of the real robot based on the data and an actual target point, and uploading the corrected teaching data to the offline teaching device; Characterized in that in the data and a fourth step of estimating calculation of the correction amount.

That is, in the first step, teaching of multiple points and multiple postures is performed, a plurality of postures at which the best correction calculation result is obtained in the logical robot model to be corrected are determined, and the teaching data is used in the second step. To download to the real robot. In the next third step, the teaching data is corrected based on the difference between the operating point of the real robot based on the teaching data and the actual target point, and the corrected teaching data is uploaded to the offline teaching device. Then, the correction amount is estimated and calculated using the uploaded data, and the correction amount of the real robot at the site is reflected on the logical robot model on the offline teaching device.

In this case, in the first step, a multi-point multi-position teaching for obtaining the best correction calculation result with the logical robot model to be corrected is performed by the offline teaching device, and the teaching data to be downloaded to the real robot is obtained. Therefore, it is possible to stably acquire teaching data from which the correction amount can be accurately estimated as compared with the conventional teaching data relying on the judgment of the operator.

As a result, it is not necessary to consider the posture at the site, and only the simple alignment is required, so that the number of work steps using the real robot can be reduced.

[0013] In the method, the first step includes: an error model creating step of creating an error model according to a model of the real robot; a teaching step of performing multi-point multi-position teaching with an offline teaching device; An estimating calculation step of estimating a correction amount using the teaching data obtained in the teaching step; and a comparing step of comparing the calculation result obtained in the estimating calculation step with the error model obtained in the error model creating step. And a determination step of returning to the teaching step if the comparison result in the comparison step does not satisfy a predetermined condition, and terminating the first step if the predetermined condition is satisfied. .

Thus, the error model obtained in the error model creating step and the calculation result obtained by estimating and calculating the teaching data obtained in the teaching step are compared in the determination step, and the comparison result is determined by a predetermined method. When the condition is satisfied, the teaching data will be downloaded in the next second step.

On the other hand, if the result of the comparison in the determination step does not satisfy the predetermined condition, the process returns to the teaching step again to perform another multi-point multi-position teaching to obtain new teaching data. The estimation calculation of the correction amount and the comparison process with the error model are performed again based on the teaching data.

Then, the above series of operations is repeated,
Teaching data of multi-point and multi-posture with which the best correction calculation result is obtained is obtained.

In the determination step, the determination may be made based on a deviation amount of the estimation model when the correction amount obtained in the estimation calculation step is reflected on the teaching data in the teaching step.

In this case, it is preferable to make the determination based on the amount of deviation between the estimation model and the error model.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an embodiment in which an off-line teaching method according to the present invention is applied to, for example, an off-line teaching system of a welding gun robot (hereinafter simply referred to as an off-line teaching system according to the embodiment) will be described. This will be described with reference to FIGS.

As shown in FIG. 1, the offline teaching system 10 according to the present embodiment is connected to a key input device such as a keyboard, a pointing device such as a mouse, etc., and monitors a logical robot model imitating a real robot. Teaching device 14 for performing offline teaching by displaying on a screen of a robot, and a robot controller 1 for controlling the real robot 16.
8.

The offline teaching device 14 has a flexible disk drive (hereinafter simply referred to as an FDD).
20) is connected, and the teaching data created by the offline teaching device 14 is FDD.
Teaching data or the like recorded on the flexible disk 22 via the FDD 20 or the like is read into the offline teaching device 14 via the FDD 20.

The robot controller 18 also has an FD
D24 is connected and the robot controller 18
The teaching data corrected and added by the FDD 24 is recorded on the flexible disk 22 via the FDD 24, or the teaching data or the like recorded on the flexible disk 22 is recorded on the robot controller 18 via the FDD 24.
Is to be read.

Further, the offline teaching device 14
Is an input device 30, such as a keyboard, as shown in FIG.
A coordinate input device 32 (pointing device) such as a mouse or a joystick, a hard disk drive (HD)
D) 34 and the FDD 20 are interface (hereinafter simply referred to as I / F) circuits 36, 38, 40 and 42, respectively.
The offline teaching device 14 is connected to a LAN used to transfer teaching data and the like in another offline teaching device.
Is connected via an I / F circuit 44 and has the monitor 12 for displaying teaching data taken in through the LAN and teaching data created by the offline teaching device 14.

This off-line teaching device 14
Operation RAM 46 used for the operation of various programs (teaching processing programs, etc.), and data storing data from external devices (LAN, coordinate input device 32, HDD 34, etc.), data processed by various programs, and the like. RAM 48, input / output port 50 for inputting and outputting data to and from external devices, and CPU (control device and logical operation device) 5 for controlling these various circuits
2 are provided.

The above-mentioned various circuits exchange data between the respective circuits via a data bus 54 derived from the CPU 52, and further via an address bus and control bus (both not shown) derived from the CPU 52. Each CPU 52
It is configured to be controlled by.

The off-line teaching system 10 according to the present embodiment performs teaching for correcting the robot model according to the procedure shown in FIG.

That is, in the first step S1, an error model corresponding to the model of the real robot 16 is created. This creation process is performed using the offline teaching device 14 or another computer.

The procedure for creating the error model will be specifically described. First, in order to simplify the explanation, the number of axes (the number of joints) of the logical robot model corresponding to the real robot 16 is assumed to be six. Therefore, the tool offset is not considered, and the error is only the zero offset error of the sensor. Here, the sensor is an encoder attached to each axis. In the following description, the real robot 16
When describing as a general robot without distinguishing between and a logical robot model, it is simply described as “robot”.

Then, the robot tool tip point (TC
When P) is adjusted to an arbitrary point, X = f _X (θ ₁ + ε ₁ , θ ₂ + ε ₂ ,..., Θ ₆ + ε ₆ ) Y = f _Y (θ ₁ + ε ₁ , θ ₂₎ + Ε ₂ ,..., Θ ₆ + ε ₆ ) (1) Z = f _Z (θ ₁ + ε ₁ , θ ₂ + ε ₂ ,..., Θ ₆ + ε ₆ ) holds.

Here, f _x , f _y , and f _z are functions determined by the mechanical structure of the robot (for example, the type of a pair or the length of a link, etc.). θ _i (i = 1, 2,..., 6) is the angle of the i-th axis (i = 1, 2,..., 6) calculated from the data detected by the axis sensor (encoder), and ε _i is the i-th axis This is the sensor offset error.

In general, when the error is considered to be a very small amount, the tip position (TCP) of the robot can be expressed by a linear expression of the error. Also, when the angle is small, the simultaneous equations of the trigonometric functions are strict equations, but generally cannot be solved. Therefore, when the Taylor expansion is performed for linearization, the robot tip position (TCP) can be finally expressed by a linear expression of the error. In this case, it can be solved because it can be brought into the solution of a general simultaneous equation.

Therefore, when the above equation (1) is arranged by ignoring the second-order and higher-order terms by regarding ε _i as a small amount, the following equation (2) is obtained.

[0033]

(Equation 1)

The above equation (2) is a simultaneous linear equation, and can be solved if independent equations equal to or greater than the number of unknowns are established. Factor _{C 0x, C oy, C 0z} , C iX, C iY and C _iZ is determined from the angle and the robot structure to be calculated from the data detected by the axial sensor (e.g., the length of the kinematic pair type and links, etc.) It can be obtained from an arithmetic expression.

When the above equation (2) is expressed in a matrix form,

[0036]

(Equation 2)

Is as follows. This is the error model.

Next, in step S2, teaching in multiple points and multiple positions is performed by the offline teaching device 14.

Usually, since three equations stand for one posture,
What is necessary is to teach only the number of postures in which the necessary number of expressions stand in the expression (3). In the above equation (3), since the number of unknowns is nine, teaching in three or more postures may be performed.

By solving the simultaneous linear equations as described above, ε ₁
Ε ₆ , X, Y, and Z are obtained, and an error model in the selected, for example, three postures is obtained. For example, if any one of ε _{1 to} ε ₆ deviates from the error range, it converges to an optimum value using the Newton method.

A large number of data relating to various postures for solving the equation (2) are prepared in advance by the offline teaching device 14 or another computer and stored in a hard disk or other external storage device as a file. At the time of the calculation, the algorithm may be assembled so as to automatically read out the data that satisfies the specific condition from the file sequentially or only by reading out the expression (2). In addition, it may be performed manually.

Next, in step S3, a correction amount estimation calculation is performed using the error model obtained in step S2. This correction amount is used to predict how much ε _{1 to} ε ₆ should be changed when the real robot 16 is taught, and this prediction greatly simplifies the on-site alignment. It will be.

The estimation calculation in step S3 is performed by using ε _{1 to} ε
₆ by substituting an appropriate numerical value or changing the values of ε ₁ to ε ₆ as appropriate in consideration of the operating environment (space, operating time, etc.) of the site, to perform the position (X, Y, Z) of the TCP. ). That is, the noise components are included in ε _{1 to} ε ₆ .

In the next step S4, the estimation calculation result obtained in step S3 is compared with the error model obtained in step S2. Specifically, the amount of deviation between the position of the TCP based on the teaching of the three postures selected in step S2 and the positions of the plurality of TCPs obtained by the estimation calculation in step S3 is determined. The amount of deviation between each of the plurality of TCP positions obtained by the estimation calculation in step S3 and another specific point is obtained.

In the next step S5, it is determined whether or not the result of the estimation calculation and the error model have become smaller than the reference value. Specifically, one or a combination of two or more of the following evaluation methods can be adopted. (1) Whether the average value of each shift amount obtained in step S4 is equal to or less than a reference value. (2) Whether the worst value among the deviation amounts obtained in step S4 is equal to or less than a reference value. (3) Whether the worst value among the deviation amounts from the specific point obtained in step S4 is equal to or less than the reference value. (4) Whether the values of the matrix elements in the arithmetic processing can be calculated. For example, it indicates whether the matrix is a regular matrix or a singular matrix, or is controllable or observable.

If an affirmative determination is made in step S5, the process proceeds to the next step S6, and if a negative determination is made, the process returns to step S2 to perform the processes in and after step S2. That is, in step S2, another multi-point, multi-posture teaching (teaching in another three postures) is performed to obtain a new error model, and in the next step S3, the new error model is obtained. The estimation calculation of the correction amount is performed again on the basis of the result, and a comparison process between the estimation calculation result and the new error model is performed in step S4.

Then, the above series of processing is repeated until a positive result is obtained in step S5. Therefore, the teaching data constituting the error model that has been affirmed in step S5 is the teaching data of the multi-point and multi-posture from which the best correction calculation result is obtained.

Next, in step S6, the teaching data of the multi-point multi-position is downloaded to the robot controller 18 of the real robot 16. This download is performed, for example, by recording the multi-point multi-position teaching data on the flexible disk 22 via the FDD 20 connected to the offline teaching device 14, and then recording the multi-point multi-position teaching data on the flexible disk 22. This is performed by causing the robot controller 18 to read multi-position teaching data via the FDD 24 connected to the robot controller 18.

In the above example, the flexible disk 22
However, if the optical disc such as an MO or a CD-R can be used as a download medium, and the offline teaching device 14 and the robot controller 18 are connected by a LAN, , May be downloaded through a LAN.

Next, in step S7, the real robot 16 is operated under the control of the robot controller 18, and the TCP is set at several target points P1, P2, P3, P
4 and P5. When the position of the target point determined from the teaching data is defined as a work point, in step S7, the work point and the actual target points P1, P2, P3, P4, and P5 are aligned. At this time, the teaching data is corrected and added by reflecting the movement amount due to the positioning on the teaching data.

The multi-point multi-position teaching data downloaded to the robot controller 18 is high-precision teaching data as close as possible to the real robot 16 using a logical robot model. , Can be performed with almost no change in posture.

Next, in step S8, the corrected teaching data registered in the robot controller 18 is uploaded to the offline teaching device 14. This uploading, for example, records the corrected teaching data on the flexible disk 22 via the FDD 24 connected to the robot controller 18 in the same manner as the download described above, and then records the teaching data on the flexible disk 22. The teaching data after the correction is input to the offline teaching device 1.
This is performed by reading the data into the offline teaching device 14 via the FDD 20 connected to the offline teaching device 4.

In the above example, the flexible disk 22
Is used, an optical disk such as an MO or a CD-R may be used as a medium for uploading, and if the offline teaching device 14 and the robot controller 18 are connected via a LAN, , May be uploaded through a LAN.

Then, in the next step S9, a correction amount estimation calculation is performed using the corrected teaching data uploaded to the offline teaching device 14. By this estimation calculation, the correction amount of the real robot 16 at the site is reflected on the logical robot model on the offline teaching device 14.

As a result, when it is necessary to perform another operation (operation) on the same real robot 16, a teaching program for performing the operation (operation) can be simplified by using the logical robot model. , And can be created with high accuracy.

In the offline teaching system 10 according to the present embodiment, the error model created in step S1 and the calculation result obtained by estimating the teaching data obtained in step S2 in step S3 are as follows. In step S4, when the comparison result satisfies a predetermined condition, the teaching data is downloaded to the robot controller 18 in step S6.

On the other hand, if the result of the comparison in step S4 does not satisfy the predetermined condition, the flow returns to step S2 to perform another multi-point multi-position teaching to obtain new teaching data. Based on the above, the estimation calculation of the correction amount and the comparison process with the error model are performed again.

Then, the above series of operations is repeated,
The teaching data of the multi-point and multi-position where the best correction calculation result is obtained is obtained.

As described above, in the offline teaching system 10 according to the present embodiment, in steps S1 to S5, the teaching of the multi-point and multi-position in which the best correction calculation result is obtained for the logical robot model to be corrected is performed in the offline mode. Since the teaching device 14 obtains teaching data to be downloaded to the real robot 16, the correction amount can be estimated more accurately than conventional teaching data that relies on the individual judgment of the operator. It is possible to obtain stable teaching data stably.

As a result, it is not necessary to consider the posture at the site, and only the simple alignment is required. Therefore, the number of work steps using the real robot 16 can be reduced.

In the above-described embodiment, the estimation calculation of the correction amount is performed by assuming a six-axis robot.
Of course, it can be easily applied to robots other than axes.

In the above-described embodiment, an example in which the present invention is applied to the off-line teaching system 10 for a welding gun robot is shown. However, the present invention can also be applied to various production robots.

The off-line teaching method according to the present invention is not limited to the above-described embodiment, but may adopt various configurations without departing from the gist of the present invention.

[0064]

As described above, according to the off-line teaching method according to the present invention, it is possible to stably acquire teaching data from which a correction amount can be accurately estimated, and reduce the number of work steps using a real robot. Can be achieved.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing an offline teaching system according to the present embodiment.

FIG. 2 is a block diagram showing a configuration of an offline teaching device in the offline teaching system according to the present embodiment.

FIG. 3 is a flowchart showing an operation of the offline teaching system according to the present embodiment.

[Explanation of symbols]

10 Off-line teaching system 12 Monitor 14 Off-line teaching device 16 Real robot 18 Robot controller 22 Flexible disk

Claims (4)

[Claims] 1. An off-line teaching method for correcting a robot model in an off-line teaching system, wherein an off-line teaching device performs multi-point, multi-position teaching using a logical robot model to be corrected to obtain the best correction calculation result. A second step of downloading the teaching data obtained in the first step from the offline teaching device to the real robot to be corrected; an operating point and an actual target point of the real robot based on the teaching data A third step of correcting the teaching data based on the difference between the teaching data and a fourth step of uploading the corrected teaching data to the offline teaching device and performing an estimation calculation of a correction amount using the uploaded data. Have An offline teaching method characterized by: 2. The offline teaching method according to claim 1, wherein the first step includes: an error model creating step of creating an error model corresponding to a model of the real robot; and a multi-point multi-position teaching using an offline teaching device. A teaching step to be performed; an estimation calculation step of performing an estimation calculation of a correction amount using the teaching data obtained in the teaching step; and a calculation result obtained in the estimation calculation step and an error obtained in the error model creation step. A comparison step of comparing models; and a determination step of returning to the teaching step if the comparison result in the comparison step does not satisfy a predetermined condition, and terminating the first step if the predetermined condition is satisfied. And an off-line teaching method. 3. The offline teaching method according to claim 2, wherein the determining step is based on a deviation amount of the estimation model when the correction amount estimated in the estimation calculation step is reflected on the teaching data in the teaching step. An off-line teaching method characterized by making a judgment by using 4. The offline teaching method according to claim 3, wherein said determining step makes a determination based on a deviation amount between said estimation model and said error model.